摘要

Advanced engine control, such as active control, often involves controlling engine with more control variables at a wider frequency range. When the conventional successive-loop-closure technology is applied to design controller with more objectives and controls, it will lead to multi-loop interleaving, and prevent the responses of the lower frequency control loops from meeting the requirements and hence exploiting the potential of the engine. A modified multivariable decoupling method in frequency domain is presented to improve the robustness and decoupling of the system and proved that it is equivalent to Rosenbrock's modified method at a single frequency, and then extended to a weighted formulation on a certain frequency range. Finally, the modified method is used to design steady-state decoupling controller, and a multivariable gain -scheduling controller for large transient is formulated as the weighted function of a family of designed decoupling controllers, whose weights are chosen as transfer function of neural network optimized by BP algorithm. Computational results are presented for controlling the small and large transient of a military aircraft turbofan engine, and the simulations show that compared with standard Hawkins's method, the resultant compensated system designed by the modified method is of better good decoupling and robustness.